Currently, solution providers have proposed deploying network monitoring devices to sniff packets on traffic flows in communication networks and perform analysis based on sniffed packets. In existing proposals, the network monitoring device (NPM) usually has the following functions: (1) receiving packets from directional traffic flows through linking to high-speed transmission links in a data network; (2) performing statistical data calculations on received packets of traffic flows; and (3) reporting calculated statistical data to an external processing device or system, such as a customer experience management (CEM) system. However, there are significant shortcomings associated with these proposals as explained below.
Whilst it is proposed that the NPM sniff and analyse packets in order to acquire statistical data relating to user traffic, communication and data network operators are interested in knowing how end users (customers) of their networks perceive throughputs (download throughput and/or upload throughput). For example, communication and data network operators are eager to determine a mean throughput in uplink/downlink of an end user (customer), where the mean uplink throughput may be determined to be a number of bits/bytes transferred over a preconfigured report period divided by the preconfigured report period.
However, current methods for determining mean throughput values suffer from inaccuracies in that the methods do not correct for packets arriving in bursts, nor IP packets transferred solely for maintaining the session between the end users' device and internet servers.
Different approaches have been proposed to address the inaccuracies. One such approach involves calculating the number of bytes transferred in the downlink/uplink direction during a confined period and then disregarding the lowest 20%-30% values. An alternative approach disregards the highest 10% values. However, both approaches may fail to reflect congestion in the network such that the network operator is unable to quickly learn of any network congestion by looking at the estimated throughput of users.
In this context, there is a need for a method which is able to accurately calculate the estimated throughput of a subscriber in a communication network.